Abstract
Stationary memoryless sources produce two correlated random sequences Xn and Yn. A guesser seeks to recover Xn in two stages, by first guessing Yn and then Xn. The contributions of this work are twofold: (1) We characterize the least achievable exponential growth rate (in n) of any positive ρ-th moment of the total number of guesses when Yn is obtained by applying a deterministic function f component-wise to Xn. We prove that, depending on f, the least exponential growth rate in the two-stage setup is lower than when guessing Xn directly. We further propose a simple Huffman code-based construction of a function f that is a viable candidate for the minimization of the least exponential growth rate in the two-stage guessing setup. (2) We characterize the least achievable exponential growth rate of the ρ-th moment of the total number of guesses required to recover Xn when Stage 1 need not end with a correct guess of Yn and without assumptions on the stationary memoryless sources producing Xn and Yn.
Original language | English |
---|---|
Article number | 159 |
Journal | Information (Switzerland) |
Volume | 12 |
Issue number | 4 |
DOIs | |
State | Published - 2021 |
Keywords
- Arimoto–Rényi conditional entropy
- Guessing
- Majorization
- Method of types
- Ranking function
- Rényi entropy
- Schur concavity
- Shannon entropy
All Science Journal Classification (ASJC) codes
- Information Systems